Conditions under which measurement-driven instruction might produce its expected results are discussed, with reference to test stakes, performance standards, and the nature of content assessment. Under some conditions, instruction driven by measurement may corrupt the measurement process. (SLD)